OBJECTIVES: When modeling long-term costs and health effects using Markov models, choosing the time of transition to another state (progression of the disease) seems to influence the final results. Various approaches can be adopted, i.e. transitions at the beginning, at the end or in the middle of the cycle. Our aim is to measure influence of cycle length and progression rates on differences between final results obtained using those methods and to establish whether there is an optimal cycle length for which half cycle correction (HCC) should always be applied. METHODS: A simple, hypothetical, two-state Markov model was built. The time horizon was set to be lifetime, an outcome discount rate was 0% or 5% and costs/utilities were held constant in time. Assuming different progression rates (0.05–0.90 annually), three methods were compared: transitions at the beginning of the cycle (‘beginning’), at the end of the cycle (‘end’), in the middle of the cycle (HCC). For each parameter the threshold values were determined, i.e. the maximal cycle length for which the difference between half-cycle correction and ’beginning’/‘end’ methods were not greater than 5%. We assumed that cycles longer than the estimated threshold will imply the application of HCC. RESULTS: For 5% discount rate the threshold cycle length was 1 year for annual progression of 0.05 and it became shorter for lower progression rates (2 weeks for 0.90 progression rate). Assuming no discounting, the threshold was 2 years for annual progression of 0.05 and 2 weeks for progression of 0.90. CONCLUSIONS: Choice of the time of transitions in the model may have a significant impact on results. For cycles shorter than 2 weeks HCC does not seem to be necessary, however it should always be applied for cycles longer than 1 year. For cycles between 2 weeks and 1 year a general recommendation cannot be made.